The influence of context on language alternation practices in English as a lingua franca

3.1 Context in ELF research

In empirical ELF research, it has become a common practice for researchers to describe context from the perspective of external contextual parameters. For example, BELF research (ELF in business contexts) draws its data from different kinds of companies, business negotiations, or simulated student interactions from business universities (e. g. Kankaanranta and Louhiala-Salminen 2018), whereas ELF in academic contexts commonly looks at different kinds of settings that physically take place at universities (e. g. Mauranen 2012). That is to say, the named contexts of these data are the external physical contexts, the geographical locations, or the topic of the interactions which brings the interlocutors together, rather than, e. g., the activity type (greetings, transactional talk, etc.). This practice likely derives from practical reasons of data collection, but the existing ELF corpora and data sets would usually allow for more detailed descriptions of contextual specifics, which would likely sharpen analyses. Recently, the ELF paradigm has become more cognizant of the influence of context on social interaction, especially concerning speakers’ pragmatic choices. For example, Baird, Baker and Kitazawa draw on emergentism and complexity theory in arguing that “any treatment of language that neglects the dynamic and contextual nature of communication is likely to misrepresent both the data gathered and the explanations for what is observed” (Baird et al. 2014: 177), thus calling for researchers to observe language through the lens of contextualization rather than isolation and compartmentalization. Canagarajah (2018) also criticizes ELF research for relying on impressionistic understandings of context that are seldom problematized or examined critically. He calls for a more expansive view of context as layered, limitless, constantly evolving, and co-constructed through language and semiotic resources. He argues that ELF study should take into consideration that “the systems or factors that mediate and shape multilingual interaction can belong to different scales of consideration, nested or overlapping with one another” (Canagarajah 2018: 4).

Yet, how may the complexities of ever-changing, layered, and multiple context be examined? Is it possible to break them into smaller parts and examine those fractions and their influence on communication? This assumption is at least entertained in this article. It is by no means argued here that context could be exhaustively described by any available empirical methods, but it must be possible to identify at least some outlines of layers within which contextualization emerges in an interactive moment. I take Mauranen’s (2012, 2013 definition of context as a starting point and broaden it through an empirical analysis. Mauranen defines context simply as “the environment that an object is embedded in or part of; in the case of language use, the two most relevant contexts are the social environment and the linguistic environment” (Mauranen 2013: 227). Her argument is that the social environment is layered from societal to the micro-level between the individuals, and these levels coexist and create a dynamic whole consisting of “the societal or macrosocial, the individual or cognitive, and the interindividual microsocial level” (Mauranen 2013: 227). Based on the analysis, the current paper will also distinguish the individual (cognitive) and interpersonal (microsocial) levels of context, but it also argues that in multi-party interactions, a group-related level can be distinguished. I will also argue that “discourse-environmental level” describes the outermost layer better than “societal level” or “linguistic environment” when an emic perspective is adopted, because these aspects may be derived from the societal context but they also have to do with other matters such as the activity type and topic of the interaction. The contextual levels are here interpreted from an emic perspective of the participants, as far as the available data allow it, and are by no means considered exhaustive of all factors that contextualize any interactive moment at any given point.

4.1 External contextual characteristics of the data sets

Section 5.1 of the analysis focuses on the ELFA corpus, a one-million-word corpus of transcribed spoken ELF discourse (approx. 131 h) recorded in conferences, seminars, lectures, and doctoral defenses at four Finnish universities. As described on the website (ELFA 2008), the corpus contains 67 % of dia-/polylogic events (discussions) and 33 % of monologic events (lectures, presentations) from approximately 650 different speakers of 51 different L1s. Notably, the exchanges included in the corpus are multi-party interactions: even in monologic events, there are always a number of interlocutors present as an audience.

Section 5.2 focuses on doctoral defense presentations and defense discussions in the ELFA corpus, totaling 22 files comprising little over 27 h of talk.^[3] This “context” is a subset of the data analyzed in 5.1 and defined by ritual-like parameters which govern the activity type, participant roles and partly also linguistic choices. Doctoral defenses at Finnish universities are formal occasions that have certain ceremonious norms concerning dress code, forms of address, and proceedings of the event that the participants are expected to follow. First, the custos, i. e., the supervisor of the dissertation, opens the event. Then the candidate gives his/her lectio praecursoria, a (max. 20 min) talk presenting the PhD study, the methods used, and the main conclusions. The candidate then invites the opponent (or opponents, if there are two) to present their critical remarks on the study. The opponent gives a short talk of the merits of the study and then presents detailed comments/criticism, to which the candidate is expected to respond. In addition to this procedure, certain formal terms of address are predefined. For example, the University of Helsinki instructs the participants as follows:

At the public examination, the form of direct address is “Mr/Madam Opponent”. […]

The Custos will introduce the doctoral candidate and the Opponent and will open the examination by saying, for example: “As the Custos appointed by the Faculty of …., I declare this public examination open.” […]

The introductory lecture may begin, for example, with the following words: “Mr/Madam Custos, Mr/Madam Opponent, ladies and gentlemen”. […]

After the introductory lecture, the candidate will turn to the Opponent and will say: “Mr/Madam Opponent/Professor/Dr NN, I now call upon you to present your critical comments on my dissertation.” […]

After thanking the Opponent, the doctoral candidate will ask the audience to make comments and pose questions: “If anyone present wishes to make any comments concerning my dissertation, please ask the Custos for the floor.”

(University of Helsinki 2018)

Doctoral defenses are open to the public, so anyone can attend these events, although usually the attendees comprise the candidate’s family and friends, faculty members, and fellow PhD students. Importantly, the speakers have little chance of knowing the extent of the linguistic repertoire of the audience. The ELFA corpus only contains recordings from defenses mainly carried out in English.

Section 5.3 concentrates on approximately 30 h of naturally occurring conversations audio recorded by multilingual couples living in Finland, Norway, the UK, Canada and Zambia collected in 2012–2018. At the beginning of the data collection, all couples identified as using L2 English between themselves. English is the main language of the interactions recorded. Sometimes children are also present. It is worth noting that these speakers know each other very well, and the majority have at least attempted to learn each other’s L1s. All except one couple were residing in either partner’s native country.

5.1 Academic ELF

Hynninen et al. (2017) find that language alternations in the ELFA corpus are commonly rather short, mostly single words (see Turunen 2012), and normally flagged, hedged and/or explicated by using a range of discourse acts. These include:

• Explicating, where a gloss, an explanation, a label and/or a clarification of meaning is added (e. g. “people first network website” in example [a] below); or the language alternation itself can act as a clarification for an ambiguous English term/item (e. g. “the, super rich nations the superreichen” [‘super rich’], Hynninen et al. 2017: 103).

• Requesting for help, where the speaker searches for the right word either by specifically asking for help or maintaining the floor during internal word-search. Often these are realized through how/what – say/call constructions (e. g. “how do you say er harjoitystyö” [‘practical assignment’], Hynninen et al. 2017: 105).

• Contextualizing, where the speaker adds a country or language, geographic location or linguistic categorization explaining the origin/purpose of the word/phrase. Often these are used together with different kinds of say/call constructions (“as they say” and “what we call” in examples [a] and [b] respectively, see below).

• Hedging with pragmatic markers such as “sort of”, “kind of” and “like” (example [b] below); general extenders “or something like that”, “or whatever”, “and so on”, “and all that” (example [c] below); discourse markers “I mean”, “you know”, “you understand” (example [c] below); the definite article “the” (where it would not be necessary grammatically; example [c] below, see also Poplack et al. 1989) and demonstratives “this”, “these” and “those” (examples [a] and [c] below).

Together, these form a pool of resources that are often combined in junction with language alternation as well as with other words or phrases that the speaker seems to consider possibly problematic to comprehend. See the following examples from Hynninen et al. (2017: 109):

1. from the main page of this people first network website pipolfastaem as they say in solomon islands pidgin ^[4]

2. we have a sort of mhm mhm mhm well what we call er sektorsforskningens avveckling the sort of taking away of sectorial research and the academisation of that research

3. want them to embrace this you know the eduskunta the or whatever the government

Although some unflagged language alternation (cognitive processing slips; Mauranen 2013) also occurs in the ELFA corpus, these are often either accompanied by hesitation markers (Mauranen 2013: 238) or speakers self-correct their slips, such as in these two examples where a Danish L1 speaker repairs men to mean ‘but’ (7.15) and a Finnish L1 speaker corrects eli to ‘in other words’ (Hynninen et al. 2017: 118):

1. S2: the reason might be we don’t know it exactly men the reason but the reason might be that in Denmark

2. S7: it was a good example of one organisatorial model of how to improve PFM eli er in other words participatory forest management

These findings show that generally speaking, in the academic lingua franca data recorded in the ELFA corpus, language alternation is considered potentially problematic in terms of its acceptability and/or intelligibility, and hence it will often not appear without some form of flagging or repair. This is explained not only by the institutional environment in which the interactions take place, but by the speakers in these different groupings not necessarily having overview of the range of their shared repertoire. Also, a social hierarchy denotes teaching-related interactional situations: while a part of the data involves peer sessions between students or at conferences, a large part of the data are asymmetrical events with a clear expert-novice distinction between the speakers. Interestingly, in all speech events in the corpus that include over 30 language alternations, the switched words/phrases regard the topic of the conversation, whether it regards the German concept bildung (in behavioral studies), gendered expressions in different languages (in women’s studies), or certain Estonian discourse particles (in social sciences).

5.2 PhD defenses

In the 27 h of PhD defense data, only 31 occurrences were tagged as language alternation by the ELFA transcription team. This already shows how uncommon mixing is in this formal academic setting. Furthermore, upon closer examination it becomes apparent that not even all the tagged items are in fact language alternation, as there are proper names without an English equivalent (see Table 1).^[5] These include: a German newspaper (Der Spiegel), two Nordic organizations (Mellemfolkeligt Samvirke, Redd Barna), and a Finnish university known by its Swedish name (Åbo Akademi). Although Redd Barna is the only name that has an English translation (Save the Children Norway), a closer examination of the recording reveals that the speaker is indeed referring specifically to the Norwegian organization widely known by its Norwegian name. There is also one transcriber mistake: a native Spanish/Catalan speaker is saying “aha” rather than “oho” (which is a Finnish token of surprise). Hence, a more accurate number of language alternations in the data is 26.

As can be seen in Table 1, the majority of language alternations in this setting are context-related terminology specific to the activity of PhD defenses. All three occurrences of doctorant are produced by one speaker, an opponent who is a native Norwegian speaker. In Norway, the official term used for the PhD candidate in doctoral defenses is doktorand. The speaker produces the word as if it were an English word, and with a slight /t/ sound in the end, which justifies the way the word is transcribed in the corpus. There is no flagging around it, which suggests that the speaker considers the word appropriate for the context and comprehensible for the audience. The other activity-induced term, the most frequent item tagged as <FOREIGN>, is kustos (or Lat. custos). It is used by several speakers from different language backgrounds, typically either in the beginning or at the end of the defense event. The occurrences of this item have been organized in Table 1 according to their order in the event: (1) opening, (2) beginning of presentation, (3) beginning of opponent’s comments, (4) thanks and asking for comments, (5) concluding the event. The use of the word follows the recommended phrasings introduced in Section 4.1 quite literally – for example, there are several modifications of the phrase “ask the Custos for the floor”. Most of these alternations are related to the function of addressing, and a relatively high number of these occurrences have been marked as reading aloud in the corpus (indicated with an asterisk in the table), which points to this word choice being preassigned for particular speech acts.

The remaining few language alternations are two side remarks between two Finnish speakers: sen tohon pöydälle (‘it there on the table’; a fraction of a speech act in which the custos tells the doctoral researcher where he placed the chalk) and sivu 34, siinä johdannossa, sivu 34 (‘page 34, in the introduction, page 34’; a similar single speech act by another custos), a slowly and clearly pronounced translation of the item “hermeneutical” which the speaker hedges with a statement of pronunciation trouble:

don’t know how to pronounce it but the hermeneutical hermeneuttinen research method

and finally a Finnish figure of speech pihtiputaan mummo (‘the granny of Pihtipudas’), which the Swedish L1 speaking Finn contextualizes with a call construction, a geographic location, and a gloss:

the the typical what you call in Finland pihtiputaan mummo in some village er the northeast Finland

In sum, the formal, ceremonious context seems to shape the language alternation practices rather drastically: on the one hand, the overall frequency of language alternation is reduced. This may also be explained by the speakers not having an overview of the repertoire of the audience. On the other hand, the alternations that occur are mainly specific, context-related terminology which are not found elsewhere in the corpus.

5.3 Family talk

In ELF couples’ conversations, the frequency and types of language alternation vary between the couples and even between the partners of a couple; however, recordings from every couple contain some language alternation. The length of these ranges from short interjections and hybrid forms through short words and phrases into entire conversational sequences. Contrary to the ELFA corpus, the language alternations are often interactionally unmarked by both speakers, blending smoothly into the surrounding talk (see Pietikäinen 2014), and especially in the newer data that includes more “family talk”, recurrent alternation back and forth between two or more languages is observed. Among the couples who have children, one of the most common indexical (albeit not necessarily flagged) types of language alternation is used to include children in the conversation. In the families where the parents are applying the one parent–one language strategy, the children appear to orient to the parents’ language switch into L1 as indicating addressee specification/inclusion in the interaction. Consider the following example, where the parents (P and J) have been discussing paperwork, when suddenly J switches into Dutch (his L1 and the language he mostly speaks to the children) as he comments on the mischiefs of the family’s puppy (for CA transcription symbols, see Appendix):

In lines 15–17, the children (C1 and C2) laugh and join the father in disciplining the dog (in Finnish, the mother’s L1 and the language the children mostly use in the data), prompted by the sudden change of language, the father’s emphatic reaction, and possibly also the topic which occurs as a contrast to the previous frame of “adult talk”.

Another fairly common connection in which ELF couples alternate languages is related to their multilingual reality: attempting to teach each other’s L1s to each other is observed in the data as an important way of bonding. For those couples who live in the native country of one partner, the native speaking partner is a central supporter in the foreign partner’s efforts to acquire the language of the society. Interlinguistic partners can translanguage across their whole repertoire when language-learning is at stake, as shown in Extract (2) which occurs after both T and L have spent a quiet moment reading:

In line 01, the L1 Belgian Dutch (Flemish) speaker T asks his Finnish wife L whether viileä translates into Dutch as wild (pronounced /velt/). Notably, here the object of interest, viileä, is flagged with a slow and clear pronunciation and word stress. Instead of using an English word as a translation, T uses the Dutch item wild. Hence all the three languages in the couple’s shared repertoire are drawn upon in this moment of teaching lexis. L then provides a translation, first an English one to the first object viileä (line 02), and then a Finnish translation to the false interpretation wild (line 04), recycling the original positioning of these words. Wild does not seem to require an English translation, so it seems to belong to the assumed shared repertoire, while both Finnish words viileä and villi are oriented to as new information for T.

ELF couples also exhibit their development as language learners of the partner’s L1 for example by reading aloud texts and eliciting commentary from the partner. They ask for the partner’s help in pronouncing words and ask for translation help. The multilingual resource is also useful during English word search, like in the following example from a discussion concerning the need to purchase new baking equipment:

In line 01, Norwegian L1-speaker K uses a hybrid of Norwegian potet (/pʊˈteːt/) and English “potato” to form a contextualization (“potato cake”, referring to Norwegian potato pancakes) in the word search for a specialized cookware item (possibly a steketakke ‘griddle’, on which the cakes are prepared). The hybrid form may be the result of the linguistic environment (Norway) where the couple reside and the topic of the equipment needed for the preparation of a culturally traditional dish, while at the same time it draws on the couple’s shared linguistic repertoire and translingual practice where Norwegian and Spanish are frequently mixed in with English.

In the next example from a Finnish-Belgian couple’s dinner-table discussion concerning the medicalization of psychiatric treatment, the word search in line 04 is performed in the wife’s L1 Finnish, while the searched-for word is inserted in the speaker’s own L1 Dutch. These means separate the word search in line 04 as a side sequence and solicit for help:

These examples show that the speakers in this context readily utilize multilingual means in the pursuit to find appropriate words and thus achieve intersubjectivity. The speakers know each other well, they have a shared history and a good grasp of the extent to which multilingual resources are shared. Pietikäinen shows how ELF couples “automatically codeswitch” place names, backchannels/response tags, and short content words “without any flagging, […] hedging, or other marking of the switch” (Pietikäinen 2014: 13), and where neither partner seems to attach interactional value to the action of language alternation – their couple lingua franca appears to include multilingual lexicon. There is also individual variation: In the ELF couple discourse data examined here, a Belgian partner characteristically replaced the question word “what” with either Dutch (his L1), wade, or Finnish (wife’s L1), mikä/mitä – interestingly, this practice was especially prominent in his talk during the few days when he had a cold. His cognitive/physical state thus seemed to affect his individual language alternation practices. A Mexican wife characteristically produced assessments (e. g. muy rico, ‘delicious’; no me gusta, ‘I don’t like it’), interjections (e. g. ay perdón, ‘oh sorry’) courtesies (e. g. gracias, ‘thank you’) in Spanish but also in Norwegian (tusen takk, ‘many thanks’), whereas her Norwegian husband often showed agreement with typically Nordic ingressive pronunciation of the response tag “yeah” (with modifications .ja: and .ha:). Commonly these switches occurred without any of the interactional flagging types that were presented in 5.1, nor were they marked by intonation, pauses, or hesitations, which suggests that they are so frequently used in this way – possibly due to a personal preference or habit – that they do not need to be flagged.

The topic and mood of the interaction also seems to induce language alternation in ELF couples’ conversations. While discussing the prospect of bee farming, the Finnish-Belgian couple used many Finnish nouns related to bee farming, e. g. mehiläispunkki (‘varroa destructor’), punkki (‘tick’), pönttö (‘artificial hive’), and mesi (‘nectar’), whereas a Mexican-Norwegian couple often used Spanish and Norwegian content words in the kitchen, e. g. mucho limón (‘a lot of lime’), sopita (‘pasta soup’), melis (‘icing sugar’), and slikkepott (‘rubber spatula’). Furthermore, the same Mexican wife C also used Spanish hyperboles and phrases to change the mood or footing of the conversation, like in Extract (5) below where her playful tone is met by her Norwegian husband K’s accusation that her tiredness is her own fault (lines 09–10). In line 14 she manages to shift the mood back to playful key (as evidenced by laughter in the lines that follow) by using childish voice and a repeated Spanish adverb mucho:

All except one couple reside in a non-English speaking environment, and everyday terminology from the surrounding linguistic environment seems to seep into their common linguistic repertoire. For example, in Finland, the sauna stove is referred to as kiuas whereas in Norway, kindergarten is recurrently referred to as barnehage, like in Extract (6) by Chinese L1 speaker C when talking to her husband about the doings of their son in a local kindergarten:

The couples also invent their own words through multilingual wordplay as in Extract (7), where M uses a French pronunciation of the Finnish word tarjous (‘offer’), which her German-Hungarian husband had accidentally come up with after moving to Finland. It had first become an insider joke between the two and then, in their close contact over time, a part of the couple’s own lexis (here used in a context of assessing the quality of cocoa powder):

In sum, the couples are very well tuned in to what extent their multilingual repertoires overlap and how it has developed over time in close contact. The lack of social distance allows them to adopt a private “translanguaging space” (Li 2018) where linguistic resources are adopted to the “couple tongue” (Pietikäinen 2017) and adapted for specific discursive purposes but also without an apparent purpose.

6.1 Individual level

Although this level is not extensively studied in the current paper, I will here outline the individual level referring to previous (mainly cognitive) research on language alternation. As already argued, an individual’s multilingual repertoire – consisting of all the resources available in languages and varieties in which the speaker may be maximally proficient or, on the other hand, only able to recognize features (e. g. Blommaert 2012; Blommaert and Backus 2013) – is of course the central resource on which language alternation is based. However, several cognitive studies suggest other triggers for language alternation that move beyond the conscious level of whether the speaker will limit their speech to one named language or let loose their whole translingual capacity. Psycholinguistic studies on cross-linguistic influence have found that the selection of “neighbors,” words that differ only in one letter or sound from the target word, make the selection process slower even when the neighbors are in different languages (Jared and Kroll 2001; van Heuven et al. 1998). The same applies to cognates between different languages (see Dijkstra and van Heuven [2002] for an overview). These results have been taken as evidence that in language processing, languages are not activated one at a time only (de Bot 2004); however, there is evidence that a speaker must use their languages actively enough before they begin to “interfere” with one another (van Hell and Dijkstra 2002). Hence the frequency of use seems to affect the varying strengths of the links between different features in the multilingual lexicon (Jarvis and Pavlenko 2008: 87–88). For example, if a speaker frequently uses a certain word from a certain language, this word in this language is more easily retrievable from their repertoire, and therefore also more readily available for translanguaging. This may partly explain some of the speakers’ individually characteristic language alternations of certain (types of) words in Section 5.3 that seemed to have no particular interactional purpose. According to Jarvis and Pavlenko, similar “ease of retrieval” also seems to occur with items that have recently been used in a particular language, which relates to our cognitive processing capacity.

In addition to repertoire, frequency and recency affects, other psychological/cognitive aspects such as lack of focus due to environmental stressors (e. g. noise), fatigue, stress, emotional load, etc., may also be factors that on the individual level result in more frequent language alternation (Dornic 1978). In the analysis of 5.3, it was observed that during the days when a Belgian husband was suffering from a cold, his language alternations (both to Dutch and Finnish) became more frequent particularly concerning repair initiation, which he used extensively, possibly due to hearing problems caused by stuffiness. However, as the current study was not performed in laboratory conditions and more detailed information on the speakers’ cognitive state was not available, it is difficult to conclude whether these alternations occurred due to a cognitive aspect, or whether, for example, they were just a matter of personal preference .

6.2 Interpersonal level

In addition to the individual level, speakers are also compelled to consider sociolinguistic aspects on the interpersonal level. One of the obvious prerequisites for translanguaging across languages is the knowledge or estimation of the range to which speakers’ multilingual repertoires overlap. An interesting earlier observation made of social dyadic ELF interactions between complete strangers (Brunner and Diemer 2018) found that the first instance of language alternation appeared to pave the way for further language alternations relating to the speakers’ linguacultural backgrounds. Unfamiliar speakers thus seemed to both test and evaluate the scope of the shared repertoire on the go, and when language alternation was approved as an appropriate communication strategy, it was more often deployed. In the analysis of 5.3, routinized language alternation and specific lexis belonging to the “couple tongue” can be observed (see e. g. Extract [7]). The close interpersonal contact over a long time between the speakers and the frequent use of multilingual practices have likely contributed to translanguaging becoming a routinized feature of the shared practice and led to the expansion of the shared range of multilingual repertoires.

I will next discuss the social distance between the speakers and, partly relatedly, recipient design , the multitude of sensitive ways in which speakers adjust their turns to orient to their interlocutor(s) (Sacks et al. 1974). Based on the observations made earlier of dyadic ELF couple interactions, it would be easy to claim that a frequent interpersonal contact opens up for more common language alternation in ELF. However, the kind of an interpersonal relationship the speakers have may actually affect whether translanguaging is deemed appropriate. Beyza Björkman analyzes ELF data from PhD supervision sessions between PhD students and their supervisors (e. g. Björkman 2016, 2018). In her data, although the supervisor and PhD student are in regular contact with one another, there is a noticeable absence of language alternation (Björkman, personal communication). Björkman’s participants all reside in Sweden, which could easily prompt switches to Swedish (see “linguistic environment” below), but this is not the case in her data. It is thus conceivable that the social distance of the speakers suppresses language alternation to an extent. Speakers may assume translanguaging inappropriate in student–professor interaction, whereas between speakers who are on the same level of social hierarchy (e. g. couples), such considerations of appropriateness are less salient. In any case, speakers’ social relations influence the way they design their turns to their interlocutors. Drawing on data from the ELFA corpus, Mauranen (2013: 235–236) shows how a seminar leader welcomes a Swiss newcomer to the group by using a German greeting. The language alternation indicates solidarity and is specifically addressed to the (assumed) German-speaking recipient. Like also seen in Sections 5.2 and 5.3, some alternations were indicative of addressee specification and used for recipient design in including and excluding speakers from the conversation, which is of course a motive for language alternation extensively discussed in earlier bilingualism research.

6.3 Group level

In addition to individual and interpersonal levels, if there are more speakers or an audience present, speakers will have to consider the group-related level of context. As has been showed by, e. g., Cogo (2012) and Kalocsai (2014), ELF CoPs develop certain kinds of translanguaging practices specific to the group. Within larger CoPs, there may also be smaller social groupings that develop their own practices (Kalocsai 2014). Not all ELF groupings are, however, CoPs in the Wengerian sense (see e. g. Mauranen 2012) – I would for example argue against calling families CoPs due to their purpose of being together being less goal oriented.^[6] There may also be temporary groupings that only come together for a short duration of time, take place physically or online, or both. As mentioned earlier, Pitzl (2018) argues that these kinds of transient international groups (TIGs) make up a multilingual resource pool which is gradually revealed and expanded through interaction. She argues that “multilingual ELF speakers could at any point during any ELF interaction decide to draw upon any of the *languages in their [individual multilingual repertoire]” (Pitzl 2018: 33, original emphasis). Often they do not, as is shown in the doctoral defense data where language alternation was scarcer than in the ELFA corpus on the whole and comprised mainly event-specific terminology. It is therefore likely that the group constellation and their shared history (or lack thereof) will also have an effect on the extent to which speakers can translanguage. The more people present and the less familiar they are with one another, the less likely everyone’s multilingual repertoires will overlap, and therefore language alternation may be perceived as a serious threat to intelligibility, such as in PhD defenses where there may be several members of the audience whose multilingual repertoire is not in any way obvious to the speaker. Also, the group hierarchy and social ranking may have an effect on the translanguaging practices that speakers adopt. In the family interactions, speakers developed specific practices related to their family unit – language alternation to address children, ritual unflagged translanguaging, and word play – while in the academic data where speakers’ relationships were largely based on professional or student roles, language alternation was less frequent and often interactionally flagged, suggesting that its acceptability needed to be negotiated in the group (Hynninen et al. 2017).

6.4 Discourse environmental level

Speakers have an innate sensitivity to the intricate normativity of social interaction, but in different lingua franca situations, these norms may not be shared. There seem to be several aspects that influence speakers’ understanding of the social situation at hand, and whether and how translanguaging is an appropriate practice. Adapting contextual aspects from Auer (2007 [1995]) on the discourse environmental level, such aspects may be the larger activity (speech genre), such as “PhD defense” where the prescriptive norms obligate speakers to use the word kustos in referring to the supervisor, the small-scale activity (speech act) such as a side remark by a supervisor within this setting that informs the PhD defendant about the location of a chalk in their shared L1 (see 5.2). The mood (key) which speakers (attempt to) create also seems to stimulate language alternation, e. g. C’s childish voice combined with language alternation to her L1 in Extract (5) manages to change the mood of the conversation from serious back to playful, while the formal undertone that describes PhD defenses in 5.2 may reduce language alternation overall. The conversation topic also sparked translanguaging as in the academic contexts related to specific disciplines or in the couples’ conversations concerning bee farming and kitchen-related talk. This may also have to do with the necessary vocabulary not being readily available for the speakers in their L2 English, so the influence of topic onto language alternation may also be linked to individual linguistic repertoires. In the ELF couples’ context, it may also just be related to the family’s linguistic environment (see below) in which the speakers have created their private “translanguaging space” (Li 2018; see also Pietikäinen 2017). In addition, the level to which the speech act has been prearranged may also affect the perceived acceptability of language alternation. In the ELFA corpus, speech acts that are typically monologous and prearranged (at least to a degree; i. e. PhD defense presentations, lectures, seminar presentations) contain fewer language alternations than the discussions that follow them (when data from translation studies is not taken into account; see Turunen [2012: 25] for detailed calculations).

Lastly, the linguistic environment where the interaction takes place has been demonstrated to influence the language alternation practices of speakers (see Pölzl and Seidlhofer 2006). As shown in 5.3, ELF couples typically adopt local terminology bound to the living environment or country in their couple tongue (e. g. barnehage, kiuas). However, the environmental effect may also be connected to the language-learning processes of the speakers: The partners in ELF relationships have invested in learning the language(s) of their environment (often the partner’s L1), and items from this language may be more readily retrievable than items from English (see recency and frequency outlined above). This highlights the dynamic interconnectedness of the different contextual levels.